fix(es/transforms/cjs): Reassign exports for the functions

[kwonoj]

• closes Object.defineProperty does not allow reassign ref to exported fn #2549.

This PR attempts to close #2549. As referenced issue(swc-project/swc-node#610) states it was not limited to Object.defineProperty, instead all of the function declarations are not assigned into exports. This makes reassigning outside of module scope does not work.

Per suggestion (#2549 (comment)) , I started looking emissions but actually it was realated to scope. I noticed this by checking test cases where exporting object actually creates exports assignment correctly:

// input 
export const someObj = { ... }

Object.defineProperty(...);

// output
Object.defineProperty(test, "errors", {
    get: ()=> someObj 
    ,
    set: (v)=>{
        exports.someObj  = someObj  = v; //current master correctly emits this if someObj is not a function
    }
});

It is due to scope includes var declaration (https://github.com/swc-project/swc/blob/master/ecmascript/transforms/module/src/common_js.rs#L210) but doesn't do it for the functions. PR attempt to do the same thing by putting fn declaration into scope as well.

One thing I'm a bit hesitant is this change uses exported_vars. While this doesn't seem to create any regressions and looks like fixes issue correctly, it may not be a good idea to mix usage of exiting property for the variable declarations.

[kdy1]

I checked it, and I think it would be better to name it exported_bindings instead of exported_vars. Can you rename it? (I checked that it has scoped visibility, so it's fine)

[kwonoj]

name it exported_bindings instead of exported_vars. Can you rename

Sounds good to me, applied changes.

[kwonoj]

Not sure about Benchmark / Performance regression check failure - peeking log, this looks like not directly related with PR's change?

[kdy1]

Thanks!